基于相空间重构的风电场日有功功率组合预测  

Combined Prediction of Daily Active Power in Wind Farm Based on Phase Space Reconstruction

作　　者：陈伟[1] 赵庆堂[1] 郭建鹏[1] 王维州[2] 肖骏[2] 

机构地区：[1]兰州理工大学电气工程与信息工程学院,兰州730050 [2]甘肃省电力公司电力科学研究院,兰州730050

出　　处：《电力系统及其自动化学报》2014年第6期20-24,36,共6页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(51267012);甘肃省电网公司科技项目(2010406029);甘肃省高等学校基本科研业务费专项资金项目(1103ZTC141)

摘　　要：风力发电具有波动性、随机性和间歇性,因此准确预测风电场的日有功功率对风电场与电力系统的稳定运行具有重要的意义。利用C-C法对风电场的日有功功率时间序列进行相空间重构,并通过计算其最大Lyapunov指数,验证了此功率时间序列具有混沌属性。在此基础上,用相空间重构建立了RBF神经网络和最小二乘支持向量机预测模型,对预测结果采用协方差优选确定权重,进行组合预测。通过对甘肃省酒泉地区某风电场的实测数据进行仿真,证明了该组合模型的有效性和可行性,并有效提高了预测精度。Wind power generation possesses the inherent features of volatility,randomness and intermittence,thus it is significant for wind farm and stable operation of power system to accurately predict the daily active power of the wind farm.Via the C-C algorithm,this paper reconstructed the phase space for the daily active power time series of wind farm and calculated the largest Lyapunov index,so as to prove that the power time series has the chaos characteristics.On the basis of chaotic theory,the phase space reconstruction was utilized in the RBF neural network and the least squares support vector machine,and the prediction results＇s weghts are calculated by covariance optimization for combined prediction.The simulation is derived from the data collected from the wind farm which is located in Jiuquan,Gansu Province,then the effectiveness and feasibility of the combined model has been proved by simulation results; furthermore,the prediction accuracy has been improved effectively.

关 键 词：风力发电 相空间重构 RBF神经网络 最小二乘支持向量机 组合预测 

分 类 号：TM614[电气工程—电力系统及自动化]